探讨机器人辅助CT引导钩丝(Hook-wire)定位在肺部磨玻璃结节(groundglass nodules, GGN)电视辅助胸腔镜手术(videoassisted thoracoscopic surgery, VATS)前的安全性与准确性，构建决策树(decision tree algorithm, DTA)预测模型。

选择2022年1月至2025年4月我院收治的经机器人辅助CT引导钩丝定位VATS手术GGN患者378例，定位成功345例为对照组，定位未成功33例为观察组。比较两组临床资料、定位准确率及穿刺相关并发症；构建DTA预测模型，采用受试者操作特征(receiver operating characteristic, ROC)曲线、校准曲线及5折交叉验证判断预测效能。

定位未成功33例(8.73%)，其中钩丝脱出肺组织并异位至胸膜腔9例，钩丝移位24例。定位相关并发症气胸27例(7.14%)，自限性肺出血46例(12.17%)；轻微胸膜反应2例；无血胸、空气栓塞等危重并发症。观察组肺出血高于对照组10例(30.30%)比36例(10.43%)(P＝0.001)。单因素分析显示，结节大小、结节到胸膜距离、总深度、吸烟史及病灶位置与定位不良相关(P<0.05)。DTA模型中特征权重结节到胸膜距离0.454、总深度0.245、肺结节大小0.217、吸烟史0.076和病灶位置0.007。ROC分析训练集曲线下面积(area under the curve, AUC)为0.909(95%CI：0.716~0.996)，验证集AUC为0.950(95%CI：0.901~1.000)；5折交叉验证后模型准确率为83.2%。校准曲线显示预测概率与实际发生率一致性良好。术后随访，肿瘤复发/转移45例，死亡19例。

机器人辅助CT引导钩丝定位用于GGN术前定位安全可靠；结节到胸膜距离、结节大小、总深度、吸烟史及病灶部位是定位失败的影响因素；DTA模型可预测术前定位失败风险。

To investigate the safety and accuracy of robotassisted CTguided hookwire localization before videoassisted thoracoscopic surgery (VATS) for pulmonary groundglass nodules (GGNs), and to construct a decision tree algorithm (DTA) prediction model.

A total of 378 patients with GGNs who underwent robotassisted CTguided hookwire localization followed by VATS in our hospital from January 2022 to April 2025 were enrolled. Among them, 345 cases with successful localization were assigned as the control group, and 33 cases with unsuccessful localization as the observation group. The clinical data, localization accuracy, and puncturerelated complications were compared between the two groups. A DTA prediction model was constructed, and its predictive performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and 5fold crossvalidation.

Localization failed in 33 cases (8.73%), including 9 cases of hookwire dislodgement from lung tissue with ectopic placement into the pleural cavity and 24 cases of hookwire displacement. Localizationrelated complications included pneumothorax in 27 cases (7.14%), selflimiting pulmonary hemorrhage in 46 cases (12.17%), and mild pleural reactions in 2 cases. No critical complications such as hemothorax or air embolism occurred. The incidence of pulmonary hemorrhage in the observation group was significantly higher than that in the control group [10 cases (30.30%) vs. 36 cases (10.43%), P=0.001]. Univariate analysis showed that nodule size, distance from nodule to pleura, total depth, smoking history, and lesion location were associated with poor localization (P<0.05). In the DTA model, the feature weights were: distance from nodule to pleura 0.454, total depth 0.245, nodule size 0.217, smoking history 0.076, and lesion location 0.007. The ROC analysis yielded an area under the curve (AUC) of 0.909 (95%CI: 0.716~0.996) for the training set and 0.950 (95%CI: 0.901~1.000) for the validation set. After 5fold crossvalidation, the model accuracy was 83.2%. The calibration curve showed good consistency between the predicted probability and the actual incidence. During postoperative followup, tumor recurrence/metastasis occurred in 45 cases, and 19 patients died.

Robotassisted CTguided hookwire localization is safe and reliable for preoperative localization of GGNs. Distance from nodule to pleura, nodule size, total depth, smoking history, and lesion location are influencing factors for localization failure. The DTA model can effectively predict the risk of preoperative localization failure.